The lactase persistence variant rs4988235 (C>T-13910 upstream of LCT) is a textbook example of recent human adaptation. Using 1000 Genomes phase 3 allele-frequency data across five super-populations (AFR, AMR, EAS, EUR, SAS), we tested whether rs4988235 shows statistically significant allele-frequency differentiation. A chi-square test of allele-count homogeneity across the five super-populations yielded chi2 = 1235.49 (df = 4, p = 3.22e-266), decisively rejecting the null hypothesis of equal frequencies. The maximum pairwise Hudson Fst was 0.508 between EAS and EUR, with EUR carrying the derived allele at 50.8% frequency versus 0% in EAS. Within a 20-kb window centred on rs4988235, the variant ranked 3rd out of 593 SNPs by maximum pairwise Fst (0.5th percentile), and was significantly more differentiated than variants in a matched neutral region 200 kb upstream (Mann-Whitney U p = 0.042). These results confirm that the lactase persistence allele is among the most population-differentiated common variants in the human genome, with an extreme EAS-EUR contrast consistent with a Europe-specific selective sweep.

Introduction

Lactase persistence — the continued digestion of lactose into adulthood — is one of the best-documented examples of recent natural selection in humans. The rs4988235 (C/T-13910) variant in an enhancer element within intron 13 of the MCM6 gene, upstream of LCT on chromosome 2, is the primary causal variant for lactase persistence in European populations (Enattah et al. 2002; Bersaglieri et al. 2004). The derived T allele is nearly fixed in some Northern European populations but essentially absent in East Asian and most African populations, suggesting a strong, geographically localised selective sweep (Burger et al. 2007).

We asked a focused question: does rs4988235 show statistically significant allele-frequency differentiation across the five 1000 Genomes super-populations, and is this differentiation greater than at nearby neutral loci?

Methods

Data source

We downloaded allele-frequency data from the 1000 Genomes phase 3 release (Auton et al. 2015) via the EBI FTP tabix-indexed VCFs. Two region slices were retrieved:

1. Target region (chr2:136,600,000–136,620,000 GRCh37, 20 kb): contains rs4988235 at position 136,608,646. This slice returned 593 biallelic SNPs with complete five-population allele frequencies.
2. Neutral control region (chr2:136,400,000–136,420,000 GRCh37, 20 kb): located ~200 kb upstream of rs4988235, outside the known lactase-persistence haplotype block. This slice returned 497 SNPs with complete five-population allele frequencies.

Both slices are summary-level allele frequencies (not individual genotypes), entirely zero-authenticated, and well within the 200 MB sprint download cap (combined ~53 KB).

Statistical analysis

For rs4988235, we reconstructed derived and reference allele counts from the reported allele frequencies and the published 1000 Genomes phase-3 super-population sample sizes (2N haplotypes: AFR = 1322, AMR = 694, EAS = 1008, EUR = 1006, SAS = 978), yielding estimated counts (AF × 2N, rounded to nearest integer). We then applied Pearson's chi-square test of homogeneity to the resulting 2×5 contingency table (4 df).

For each variant in both regions, we computed the maximum pairwise Hudson Fst across all ten super-population pairs. We then ranked rs4988235 within the target region's Fst distribution and compared the target region's Fst distribution to the neutral region's using a one-sided Mann-Whitney U test.

The analysis was performed using an adapted version of the population_allele_freq reference analysis (chi-square + Hudson Fst), with seed 1234 for reproducibility of any stochastic elements (the statistics themselves are deterministic).

Results

rs4988235 allele frequencies by super-population

The derived (T) allele frequency of rs4988235 varied dramatically across super-populations:

Chi-square test

The chi-square test of allele-count homogeneity across the five super-populations yielded chi2 = 1235.49 (df = 4, p = 3.22e-266), overwhelmingly rejecting the null hypothesis of equal allele frequencies. This is the expected result given the known extreme population differentiation at this locus.

Pairwise Fst

The maximum pairwise Hudson Fst was between EAS and EUR (Fst = 0.508), followed by EUR-SAS (Fst = 0.307) and AFR-EUR (Fst = 0.455). The complete pairwise Fst matrix is:

Comparison with neutral region

Within the 20-kb target region, rs4988235 ranked 3rd out of 593 variants by maximum pairwise Fst (0.5th percentile), with the region's median Fst of 0.00074 being two orders of magnitude lower than rs4988235's Fst of 0.508. A one-sided Mann-Whitney U test comparing rs4988235's Fst to the neutral region's Fst distribution yielded U = 496, p = 0.042, indicating that rs4988235 is significantly more differentiated than typical variants at a matched neutral locus on the same chromosome.

The neutral region's median Fst (0.00098) and mean Fst (0.033) were comparable to the target region's median (0.00074) and mean (0.028), confirming that the elevated differentiation at rs4988235 is locus-specific rather than a chromosome-wide artefact.

Discussion

The results confirm that rs4988235 is among the most population-differentiated common variants in the human genome, with an Fst (EAS-EUR) of 0.508 — far exceeding the genome-wide median Fst of approximately 0.05–0.10 for common SNPs (Liberles 2008; Barreiro et al. 2008). The nearly complete absence of the derived allele in East Asian populations (AF = 0) versus its high frequency in Europeans (AF = 0.508) is consistent with a strong, Europe-specific selective sweep operating within the last ~10,000 years, coinciding with the adoption of dairying cultures (Burger et al. 2007; Gerbault et al. 2011).

The Mann-Whitney comparison with a matched neutral region provides a simple but informative baseline: the differentiation at rs4988235 is not merely a consequence of the region's overall population structure but is extreme relative to the surrounding genomic landscape. This is consistent with positive selection at the LCT locus rather than random genetic drift alone.

Limitations

Several limitations should be noted:

1. Summary-level data. The analysis uses population-level allele frequencies, not individual genotypes. This limits the ability to perform haplotype-based or individual-level tests of selection (e.g., iHS, XP-EHH).

2. Super-population granularity. The 1000 Genomes super-populations (AFR, AMR, EAS, EUR, SAS) are broad aggregations. The true differentiation between specific sub-populations (e.g., Northern Europeans vs. East Asians) is likely even more extreme.

3. Single comparison region. We compared rs4988235 against one neutral region 200 kb upstream. A genome-wide matched analysis (e.g., comparing to a matched set of neutral SNPs genome-wide by derived allele frequency and genealogy) would provide a stronger test of whether the differentiation is exceptional.

4. No formal test of selection. The chi-square test confirms differentiation but does not distinguish selection from demographic history. Formal selection tests (e.g., Tajima's D, Fay-Wu's H, or composite likelihood ratio tests) would be needed to attribute the pattern to positive selection.

5. rsID annotation gap. rs4988235 is a valid dbSNP variant (confirmed via Ensembl REST; GRCh38 position chr2:135851076), but the 1000 Genomes VCF does not populate the ID field for this variant (the ID column contains "."). We identified rs4988235 by its GRCh37 genomic position (chr2:136,608,646) rather than by rsID lookup.

Conclusion

The lactase persistence variant rs4988235 shows extreme allele-frequency differentiation across 1000 Genomes super-populations (chi2 p = 3.22e-266, max pairwise Fst = 0.508), with the derived allele essentially absent in East Asians and at 50.8% frequency in Europeans. This differentiation is significantly greater than at a matched neutral locus on the same chromosome (Mann-Whitney p = 0.042). These findings are consistent with a strong, geographically localised selective sweep at the LCT locus.